Authentic Creation: Prove You Beat the Bots

Human creators are seeking a definitive 'AI-free' label for their work, yet a universally agreed-upon standard remains elusive.

The phrase "This looks like AI" has become a source of apprehension for many human creatives, including writers, illustrators, and photographers. As generative artificial intelligence increasingly excels at replicating human-produced content, a natural skepticism arises among consumers, particularly when online platforms fail to clearly identify even overtly AI-generated material.

This growing challenge suggests a potential solution: establishing a widely recognized label for human-made text, images, audio, and video, similar to a Fair Trade certification. While AI systems lack the incentive to identify their own creations, human creators, facing potential displacement, are strongly motivated to distinguish their authentic work.

This perspective is gaining traction within the industry. Adam Mosseri, head of Instagram, articulated a similar sentiment in December, proposing that as AI technology advances to produce content visually indistinguishable from professional human work, it will become "more practical to fingerprint real media than fake media."

While the exact volume of AI-generated content across the internet remains unquantified, a recent Reuters Institute survey indicates a widespread public perception that news sites, social media platforms, and search engine results are increasingly saturated with it.

The C2PA content credentials standard, already adopted by Meta's platforms, was designed to authenticate human-made works. However, despite broad industry backing, its implementation has largely proven ineffectual. A primary reason for this failure appears to be the strong incentive for those creating and hosting AI content to conceal its origins, driven by the potential for increased engagement, disruptive influence, and financial gain.

In response to this challenge, numerous solutions have emerged in recent years, aiming to help human creatives differentiate their work from AI-generated outputs. Yet, much like the C2PA standard, these initiatives encounter significant hurdles in achieving widespread adoption.

Currently, the landscape of AI-free labeling alternatives is fragmented, with at least a dozen different initiatives vying for recognition. Each offers varying eligibility criteria and authentication methodologies to address the core problem. Some, like the Authors Guild’s "human authored certification" for books and written works, are highly industry-specific, preventing their universal application across all creative mediums.

Conversely, broader solutions such as Proudly Human and Not by AI strive to encompass diverse creative outputs, including published text, visual art, videography, and music. However, the verification processes employed by these services often raise questions regarding their reliability, mirroring the challenges faced by AI-labeling solutions. For instance, Made by Human operates on a trust-based model, offering downloadable badges without verifying provenance. Others, like No-AI-Icon, claim to visually inspect works and utilize AI detection services, which are widely acknowledged for their unreliability.

The majority of these services currently rely on a labor-intensive approach: requiring creatives to manually present their working processes, such as sketches or written drafts, to a human auditor. While demanding in terms of effort, this method is presently considered the most reliable way to authenticate human authorship in the absence of more efficient technological alternatives.

A fundamental challenge lies in defining what precisely constitutes "human-made." With AI tools increasingly integrated into creative workflows and even promoted by creative educators, establishing a clear boundary for authentic human creation becomes complex.

Jonathan Stray, a senior scientist at the UC Berkeley Center for Human-Compatible AI, articulated this dilemma to The Verge, stating, "The problem is going to be definition and verification. Does chatting with an LLM about the idea before executing it manually count as using AI? And how could the creator prove no AI was involved?" He drew a parallel to established consumer labels like 'Organic,' which are supported by clear regulations and enforcement agencies.

Nina Beguš, a lecturer at the UC Berkeley School of Information, further emphasizes that society has already entered an era of hybrid content, which fundamentally conflicts with traditional definitions of authentic creation. She conveyed to The Verge, "Any creative output today can be touched by AI in one way or another without us being able to prove it." Beguš added that "Authorship is disintegrating into new directions, becoming more technologically enhanced and more collective. We need to revamp our creativity criteria that were made solely for humans."

One proposed human-made label, "Not by AI," attempts to navigate this definitional ambiguity. It provides various badges for creators to apply across websites, blogs, art, films, essays, books, and podcasts, contingent on at least 90 percent of the work being human-created. However, this voluntary system currently operates without independent verification of its claims.

In contrast, solutions like "Proof I Did It" leverage blockchain technology to establish a permanent, immutable record for verified human-made content and its creators. By anchoring verification on the blockchain, creators receive an unforgeable digital certificate confirming human authorship, offering a significantly more reliable method than relying on speculative AI detection software.

Thomas Beyer, an executive director at the University of California’s Rady School of Management, suggests that Web3 and blockchain technology present a robust solution by reframing the core inquiry from "does this look like AI?" to "can this account prove its human history?" As Beyer explained to The Verge, "By issuing ‘Made by Human’ tokens to verified creators, the market creates a ‘premium tier’ of art where authenticity is mathematically guaranteed." This sentiment is echoed by other experts, including Beguš, who foresee a potential increase in the value of "human and biological creativity" in the wake of widespread synthetic media.

Despite its current shortcomings, established standards such as C2PA offer a crucial element that AI-free labeling solutions urgently require: unification. Major technology companies, including Adobe, Microsoft, and Google, have committed to this standard, and AI providers are adopting it to satisfy global regulatory demands. However, when evaluating the merits of AI labeling versus verifying authentic human-made content, the latter approach appears to hold greater promise for success.

A significant number of creative professionals, even those not entirely opposed to AI tools, are understandably driven to differentiate their work from the burgeoning volume of synthetically-generated content that is saturating the industry and posing a threat to their livelihoods. While social media platforms host numerous AI advocates eager to demonstrate the technology's capabilities, a notable reluctance exists among others to disclose AI usage, particularly when financial gain or influence might be jeopardized.

Illustrative examples of this reluctance include porn actors creating perpetual digital clones of themselves, or AI influencers marketing non-existent fantasy lifestyles. Disclosing the AI origin in such cases could shatter the illusion of a genuine human experience for consumers. Similarly, scammers employing AI-generated imagery to peddle online products have no incentive to reveal their methods, a lack of concern often reflected by hosting platforms like Etsy. Moreover, individuals using generative AI to propagate discord or mischief on social media rely on their creations being perceived as real. These factors collectively explain why AI labeling, even with standards like C2PA, has struggled to gain traction.

Evidence already exists of AI-focused creators actively avoiding transparency. A prominent example is romance author Coral Hart, who informed The New York Times that she earned a six-figure sum last year by producing over 200 AI-generated novels. She deliberately omits any labels disclosing AI tool usage on her books, fearing it would "damage her business for that work" due to the "strong stigma" associated with the technology.

The prevailing disdain for synthetically-generated content is evident in its frequent dismissal as "slop," even when such works demonstrate visual, auditory, or technological impressiveness. This raises a critical question: how will providers of human-made or AI-free labels prevent their certifications from being fraudulently exploited by those who benefit from deception? Trevor Woods, CEO of Proudly Human, concedes that complete prevention